Research ArticleCancer

Targeting KRAS-dependent tumors with AZD4785, a high-affinity therapeutic antisense oligonucleotide inhibitor of KRAS

See allHide authors and affiliations

Science Translational Medicine  14 Jun 2017:
Vol. 9, Issue 394, eaal5253
DOI: 10.1126/scitranslmed.aal5253

This article has a correction. Please see:

An antisensible approach to targeting KRAS

Mutations that cause activation of the KRAS oncogene are common in human cancer, including treatment-resistant tumor types such as lung and pancreatic cancer. KRAS has also proven to be notoriously difficult to target with small molecules. To overcome this issue, Ross et al. have turned to genetic technology, demonstrating an antisense oligonucleotide–based therapy for inhibiting KRAS. The antisense oligonucleotide used in this study was chemically modified, allowing systemic delivery through subcutaneous injection and avoiding the need for a specialized delivery vehicle. The authors tested the efficacy of this therapy in multiple mouse models of non–small cell lung cancer and evaluated its safety in primates, demonstrating its potential suitability for translation to humans.

Abstract

Activating mutations in KRAS underlie the pathogenesis of up to 20% of human tumors, and KRAS is one of the most frequently mutated genes in cancer. Developing therapeutics to block KRAS activity has proven difficult, and no direct inhibitor of KRAS function has entered clinical trials. We describe the preclinical evaluation of AZD4785, a high-affinity constrained ethyl–containing therapeutic antisense oligonucleotide (ASO) targeting KRAS mRNA. AZD4785 potently and selectively depleted cellular KRAS mRNA and protein, resulting in inhibition of downstream effector pathways and antiproliferative effects selectively in KRAS mutant cells. AZD4785-mediated depletion of KRAS was not associated with feedback activation of the mitogen-activated protein kinase (MAPK) pathway, which is seen with RAS-MAPK pathway inhibitors. Systemic delivery of AZD4785 to mice bearing KRAS mutant non–small cell lung cancer cell line xenografts or patient-derived xenografts resulted in inhibition of KRAS expression in tumors and antitumor activity. The safety of this approach was demonstrated in mice and monkeys with KRAS ASOs that produced robust target knockdown in a broad set of tissues without any adverse effects. Together, these data suggest that AZD4785 is an attractive therapeutic for the treatment of KRAS-driven human cancers and warrants further development.

INTRODUCTION

RAS guanosine triphosphatases (GTPases) are one of the most commonly mutated gene families in cancer, and KRAS is the most frequently mutated isoform, with a prevalence of about 20% in all human cancers (1). There are several tumor types that exhibit a high frequency of activating mutations in KRAS, including three of the deadliest cancers: pancreatic, colorectal, and non–small cell lung cancer (NSCLC) (1, 2). KRAS functions as a molecular switch cycling between guanosine triphosphate (GTP)–bound (on) and guanosine diphosphate (GDP)–bound (off) states to affect intracellular translation of extracellular signaling through cell surface receptors. Oncogenic alleles of KRAS reduce intrinsic and GTPase activating protein–mediated GTP hydrolysis, resulting in increased GTP-bound KRAS and persistent signaling through downstream effector pathways (1, 2). Mutations of KRAS are associated with poor prognosis in several cancers, and there is a substantial body of evidence supporting the role of KRAS in the initiation and maintenance of cancer (37), indicating that it is an important therapeutic target.

Since the original discovery of KRAS as an oncogene in 1982, there have been intense efforts to develop a targeted therapeutic for KRAS mutant cancers (1, 2). Attempts at direct enzymatic inhibition of KRAS function have been largely unsuccessful due to the intrinsically high affinity of the enzyme for GTP. As a consequence, indirect approaches to inhibit mutant KRAS signaling have been pursued. The most advanced of these approaches entailed targeting the pathways involved in the posttranslational modification of KRAS (1), farnesylation of a C-terminal motif on KRAS by farnesyltransferase, a step necessary for trafficking of the protein from the cytoplasm to the inner face of the cell membrane and for effector pathway activation. Unfortunately, phase 3 studies of two independent farnesyltransferase inhibitors (FTIs), tipifarnib (8) and lonafarnib (9), failed to show benefit for patients over the standard of care in pancreatic and lung cancer, respectively. This was shown to be due to the activation of an alternative mechanism of prenylation of KRAS and NRAS in the presence of FTIs (1). Attention has also been given to inhibition of downstream effector signaling pathways, including members of the mitogen-activated protein kinase (MAPK) and phosphatidylinositol 3-kinase (PI3K) pathways, but to date, agents targeting members of these pathways have met with limited success clinically in KRAS-driven tumors (1). Combinations of agents targeting distinct RAS effector pathways are currently under clinical investigation, and it remains to be seen whether these combinations will be tolerated at doses that may provide therapeutic benefit (10, 11). Still, further indirect approaches, mainly preclinical to date, are focusing on the role of KRAS in the regulation of metabolic processes (6, 1214).

Despite the difficulty in identifying a viable direct approach to target KRAS therapeutically, KRAS mutant cancers remain an extremely high unmet clinical need disease segment, and this continues to drive the scientific community to identify approaches to inhibit this broadly important cancer target. Some progress has recently been made in generating covalent inhibitors that bind and selectively inhibit one of the KRAS mutant isoforms (KRAS G12C), a mutation predominantly found in lung cancer (1517). These early compounds have promise but, even if successful in the clinic, will only treat a subset of the overall KRAS mutant patient population. In addition, some progress has been made with delivery systems for a KRAS G12D–directed small interfering RNA (siRNA), which is currently being evaluated in locally advanced pancreatic cancer (18). Therapeutic nucleic acid–based approaches, including antisense oligonucleotides (ASOs), offer the potential to yield drugs for targets that have proven to be intractable to traditional drug modalities (19). With the ASO approach, inhibitors can be designed solely on the basis of gene sequence information, enabling the development of selective inhibitors to a wide range of target classes, including those previously regarded as undruggable. There have been several successes in the development of systemically administered ASOs in non-oncological settings (20, 21) and, recently, the U.S. Food and Drug Administration approved mipomersen (Kynamro) as the first systemically administered ASO drug. Continued efforts to improve the stability and potency of ASOs have resulted in the discovery of a next-generation class of ASOs that use 2′-4′ constrained ethyl (cEt) residues (22, 23). These next-generation cEt ASOs exhibit enhanced in vitro and in vivo potency compared to earlier ASO molecules, and recently, a cEt ASO targeting the previously undruggable target STAT3 (signal transducer and activator of transcription 3) has shown single-agent activity in early cancer studies, suggesting that this therapeutic approach holds promise (24). Here, we report the discovery of the clinical drug candidate AZD4785, a potent cEt-modified ASO inhibitor of KRAS expression, describe its primary pharmacology in preclinical cancer models, and characterize the tolerability of KRAS inhibition in nonhuman primates and in rodents.

RESULTS

AZD4785 is a potent and selective cEt-modified ASO inhibitor of human KRAS

AZD4785 (Ionis 651987) is an advanced chemistry (cEt-modified) (22, 23) KRAS ASO that is complementary to a sequence in the 3′ untranslated region (3′UTR) of KRAS mRNA and thus targets both the mutant and wild-type KRAS isoforms for ribonuclease H–mediated degradation (Fig. 1A). The in vitro evaluation of nucleic acid–based agents (siRNA or antisense) typically relies on cationic lipid–mediated transfection to deliver them into cells (2325). However, because the cEt chemistry increases the potency of ASOs (22, 23), these inhibitors can be evaluated in cancer cells growing in culture without any delivery agent (termed free-uptake conditions). The potency and selectivity of AZD4785 were initially assessed in A431 endometrioid carcinoma cells, which express wild-type KRAS and have been shown to be sensitive to free uptake of ASO in the absence of transfection or delivery reagents (24). AZD4785 selectively down-regulated KRAS mRNA with a half-maximal inhibitory concentration (IC50) of 10 nM in A431 cells without decreasing the expression of HRAS and NRAS mRNA isoforms (Fig. 1B). KRAS mRNA down-regulation correlated with a dose-dependent decrease in KRAS protein with no impact on NRAS or HRAS proteins (Fig. 1C). Systemic treatment of mice bearing established A431 tumors with unformulated AZD4785 at 16, 32, or 48 mg/kg (mpk) 5× weekly for 3 weeks reduced KRAS mRNA in the tumors in a dose-dependent manner without affecting the expression of other RAS isoforms (Fig. 1D), and immunohistochemical (IHC) analysis confirmed robust KRAS protein inhibition (Fig. 1E).

Fig. 1. Potent and selective down-regulation of KRAS mRNA and protein by AZD4785 in vitro and in vivo.

(A) Sequence and binding region of AZD4785 to the 3′UTR of KRAS mRNA transcript. ds, unmodified bases; ks, cEt-modified bases; mC, 5-methylcytosine; ORF, open reading frame. (B) Relative expression of RAS mRNA isoforms measured by qRT-PCR in A431 cells after 72 hours of treatment with AZD4785 or CTRL ASO. Expression was normalized to ACTIN and shown relative to phosphate-buffered saline (PBS). Representative data from a minimum of two independent experiments are shown. (C) Western blot analysis of RAS and vinculin protein in A431 cells after 72 hours of treatment with CTRL ASO or AZD4785. (D) Mice bearing A431 tumors were treated with PBS or with 16, 32, or 48 mpk of AZD4785 5× weekly for 3 weeks. RAS isoform expression was measured in tumors at the end of the study by qRT-PCR. Expression was normalized to ACTIN and shown relative to PBS. The graph shows individual tumor data, treatment group mean, and SE. mpk/w, mpk/week. (E) IHC analysis of KRAS protein expression in A431 tumors after AZD4785 treatment. Scale bars, 200 μm.

The sequence of AZD4785 was searched against the human transcriptome for any sequences with 0 or 1 mismatch or sequences with 2 mismatches and 14 matches in a row. The effects of AZD4785 on the expression of primary gene transcripts corresponding to those predicted off-target sites were analyzed by quantitative reverse transcription polymerase chain reaction (qRT-PCR) in AZD4785-treated human cells. All predicted coding off-target genes that could be measured in human cells had no or lower reductions in mRNA expression relative to on-target activity (KRAS mRNA knockdown; table S1). Together, these results demonstrate that the KRAS cEt ASO AZD4785 results in potent and selective down-regulation of human KRAS mRNA and protein both in vitro and in vivo without the need for any complex delivery formulation.

AZD4758 selectively inhibits the proliferation of tumor cells expressing mutant KRAS

We assessed the impact of AZD4785-mediated KRAS depletion on lung, colon, and pancreatic tumor cells. AZD4785 treatment produced dose-dependent inhibition of KRAS mRNA and protein expression in a panel of cell lines (Fig. 2 and fig. S1). Sensitivity to ASO-mediated target RNA reduction varied in the cell lines evaluated, potentially due to differences in productive and nonproductive uptake pathways in these cells, as has been described previously (24, 25). Because AZD4785 targets KRAS mRNA sequences distinct from the mutation codon sites, it was able to deplete KRAS, irrespective of the KRAS mutation (Fig. 2 and table S2). Cell lines expressing mutant KRAS have demonstrated variable dependency upon KRAS for viability in two-dimensional (2D) monolayer proliferation assays (26), which we have also observed (fig. S2). However, in anchorage-independent 3D growth assays, KRAS dependency was more clearly observed (fig. S2) (16, 27, 28). Consistent with this hypothesis, we demonstrated that AZD4785 inhibited the colony formation of a panel of KRAS mutant cancer cells (Fig. 2A and figs. S3 and S4). In contrast, the 3D growth of KRAS wild-type cancer cells was not appreciably affected by AZD4785 despite the fact the KRAS depletion was similar to the mutant cell lines (Fig. 2A and figs. S3 and S4). Consistent with depletion of functional KRAS protein, AZD4785 treatment resulted in the potent inhibition of MAPK pathway activity selectively in KRAS mutant cells, reducing the expression of the MAPK pathway downstream transcripts DUSP6 and ETV4 (Fig. 2B and fig. S1) (29). Furthermore, Western blot analysis of lysates from AZD4785-treated KRAS mutant or KRAS wild-type cancer cells confirmed down-regulation of both mutant and wild-type KRAS protein and demonstrated inhibition of downstream MAPK pathway signaling [including phospho-C-RAF proto-oncogene serine/threonine kinase (pCRAF), phospho–MAPK kinase (pMEK), phospho–extracellular signal–regulated kinase (pERK), and phospho–ribosomal protein S6 kinase 90 (pRSK90)] and PI3K pathway signaling [phospho-protein kinase B (pAKT)] selectively in the KRAS mutant cells (Fig. 2C and fig. S5). These data therefore show that AZD4785 down-regulates mutant and wild-type KRAS isoforms and has selective phenotypic effects on KRAS mutant cells in vitro.

Fig. 2. Effects of AZD4785 on proliferation and MAPK pathway signaling in KRAS mutant and wild-type cancer cells in vitro.

(A) Correlation between IC50 of KRAS down-regulation and inhibition of colony formation by AZD4785 in KRAS wild-type and mutant cell lines. NCI-H1299, NCI-H1793, and Colo201 have an IC50 of colony formation inhibition by AZD4785 of >10 μM. (B) Correlation between IC50 of KRAS and DUSP6 down-regulation by AZD4785 in KRAS wild-type and mutant cell lines. NCI-H1437, NCI-H1299, and Colo201 have an IC50 of DUSP6 down-regulation by AZD4785 of >10 μM. Mean IC50 is shown and generated from a minimum of two independent experiments. (C) Western blot analysis of cell lysates from NCI-H358 or PC9 cells after treatment for 72 hours with a range of doses of AZD4785.

Inhibition of KRAS is not limited by feedback activation of the MAPK pathway

The MAPK pathway (RAF-MEK-ERK) is a key effector pathway modulated by KRAS activity, and inhibitors of key nodes of the pathway are under development (30, 31). Therefore, we explored how AZD4785-mediated depletion of KRAS might differ from enzymatic inhibitors of the MAPK pathway by comparing the impact of AZD4785 with inhibitors of MEK1/2 (selumetinib/AZD6244/ARRY-142886) or ERK1/2 (SCH772984) on signaling in KRAS wild-type and mutant cells. Because the molecular target of ASOs is mRNA, they take longer to achieve maximal inhibition (protein depletion) compared to small-molecule enzymatic inhibitors (SMIs); therefore, cells were treated with ASOs for 72 hours and with SMIs for 24 hours before assessing the impact on downstream signaling. As expected, inhibition of MAPK pathway signaling was observed after treatment with AZD4785; however, this was selective for KRAS mutant cells, consistent with the effects on cell proliferation (Fig. 3, A and B, and figs. S1 and S5). In contrast, the activity of selumetinib and SCH772984 was less selective and inhibited proliferation and downstream MAPK pathway signaling of cells carrying either a KRAS mutation or a non-RAS mutant genotype (EGFR mutation in the PC9s) that would be predicted to activate the RAF-MEK-ERK pathway (Fig. 3, A and B, and figs. S4 and S5).

Fig. 3. Differentiation of AZD4785 from MAPK pathway inhibitors in vitro.

(A) Western blot analysis of MAPK and PI3K signaling in NCI-H358 and PC9 cells treated with AZD4785 and CTRL ASO for 72 hours or with selumetinib, SCH772984, or DMSO for 24 hours. (B) Expression of KRAS, DUSP6, and ETV4 mRNA measured by qRT-PCR in NCI-H358 and PC9 cells treated with a dose range of AZD4785 for 72 hours or with selumetinib or SCH772984 for 24 hours. Expression was normalized to GAPDH and expressed relative to PBS control. Data are from a representative experiment (n = 2) showing mean expression of technical replicates and SD. (C) Western analysis of MAPK pathway signaling in NCI-H358 and PC9 cells after monotherapy or combination treatment with AZD4785, CTRL ASO, and selumetinib. ASO treatment was done for 72 hours, and selumetinib treatment was done for 24 hours. For combination treatments, selumetinib was added 48 hours after dosing with AZD4785 for the final 24 hours of incubation. (D) NCI-H358 and PC9 cells grown in soft agar were treated with selumetinib in combination with CTRL ASO or AZD4785. Data are from a representative experiment from a minimum of two showing mean colony number and SD.

The MAPK pathway is under careful homeostatic control through negative-feedback mechanisms. For example, MEK1/2 and ERK1/2 inhibitors increase MEK1/2 phosphorylation at S218/222 in both KRAS wild-type and mutant cells by relieving upstream negative controls (30), which we also observed (Fig. 3A and fig. S5). Increases in pMEK were not observed upon AZD4785 treatment, and moreover, in some KRAS mutant cells, there was a decrease in pMEK (Fig. 3A and fig. S5), suggesting that AZD4785 limits feedback reactivation upstream of MEK1/2. To extend this observation, we assessed the impact of combining AZD4785 and selumetinib on MAPK pathway signaling and cell growth. Pretreatment with AZD4785 for 48 hours limited the selumetinib-induced increase in pMEK in NCI-H358 but not PC9 cells (Fig. 3C and fig. S6A). The combination treatment in NCI-H358 cells was also more effective at inhibiting downstream signaling than the monotherapy treatments (Fig. 3C and fig. S6A). Furthermore, selumetinib and AZD4785 combination enhanced the inhibition of cell growth in colony formation assays compared with either single-agent treatment in KRAS mutant cells, and no effect was seen with a control ASO (CTRL ASO) (Fig. 3D and fig. S6B).

Together, these data show that KRAS depletion by AZD4785 has distinct effects on MAPK signaling compared to MEK1/2 and ERK1/2 inhibitors, with preferential activity in cell lines that carry a KRAS mutation and lack of pathway reactivation through feedback phosphorylation of MEK1/2 at S218/222. Furthermore, these data support the rationale of a combination of AZD4785 with MAPK pathway inhibitors in KRAS-dependent tumors.

Systemic delivery of unformulated AZD4785 produces antitumor activity in KRAS mutant xenograft models in vivo

A pharmacokinetic (PK) and pharmacodynamic (PD) study with systemic delivery of 40 mpk of AZD4785 5× weekly in nude mice bearing NCI-H358 tumors demonstrated that maximum KRAS mRNA knockdown in tumors occurred after about 2 weeks of dosing and that this correlated with maximal tissue exposure (fig. S7). The slower onset kinetics of ASO activity is associated with accumulation of drug in tissues over time to concentrations that produce robust target engagement (32). In xenograft experiments, this is observed as a delay in antitumor activity (Figs. 4 and 5). In the NCI-H358 KRAS mutant lung cancer xenograft, treatment of animals for 4 weeks with AZD4785 resulted in robust down-regulation of KRAS mRNA (55%; P < 0.001) in tumor tissue and a reduction in downstream effector pathway activity with inhibition of DUSP4 (51%; P < 0.001) and DUSP6 (55%; P < 0.001) mRNA (Fig. 4A). As expected, AZD4785 did not down-regulate HRAS or NRAS isoforms in NCI-H358 tumors (fig. S8). Furthermore, treatment was well tolerated (fig. S9), and significant antitumor activity was observed in the NCI-H358 model, with 4 weeks of dosing of AZD4785 achieving a tumor growth inhibition (TGI) of 72% (P = 0.005) compared to PBS (Fig. 4, B and C, and fig. S10A). Significant tumor PD and efficacy were also observed with AZD4785 treatment in the NCI-H1944 KRAS G13D mutant lung cancer model (Fig. 4, D to F, and fig. S10B). Inhibition of KRAS (68%; P < 0.001), DUSP4 (73%; P < 0.001), and DUSP6 (75%; P < 0.001) mRNA was measured in tumors at the end of the study (Fig. 4D) and was associated with a TGI of 80% (P = 0.001) compared to PBS after 51 days of dosing (Fig. 4E). To assess the potential clinical translation of the antitumor activity observed, we performed Kaplan-Meier survival analyses where we used an increase in tumor size of fourfold as a surrogate for survival as an end point. In both the NCI-H358 and NCI-H1944 studies, treatment with AZD4785 delayed the time to reach the surrogate end point compared to the control groups (Fig. 4, C and F).

Fig. 4. PD and efficacy of AZD4785 in KRAS mutant lung cancer xenograft models.

(A to C) Mice bearing NCI-H358 tumors were treated with PBS, AZD4785, or CTRL ASO at 50 mpk 5× weekly for 4 weeks. (A) KRAS, DUSP4, and DUSP6 mRNA were measured in NCI-H358 tumors at the end of the study by qRT-PCR. The expression was normalized to 18S ribosomal RNA (rRNA) and expressed relative to PBS. Data are shown as individual tumor data, treatment group geometric mean, and SE. (B) AZD4785 significantly inhibited tumor growth of NCI-H358 tumors compared to PBS (TGI, 72%; P = 0.0047) and CTRL ASO (TGI, 75%; P = 0.001). Data are shown as the geometric mean of the tumor volume and SE. (C) Surrogate endpoint survival graphs for the NCI-H358 study. Data are shown as the percentage of remaining animals with tumors <4× the initial starting volume in each treatment group. (D to F) Mice bearing NCI-H1944 tumors were treated with PBS, AZD4785, or CTRL ASO at 50 mpk 5× weekly for 7 weeks. (D) KRAS, DUSP4, and DUSP6 mRNA were measured in NCI-H1944 tumors at the end of the study by qRT-PCR. The expression was normalized to ACTIN and presented relative to PBS. Data are shown as individual tumor data, treatment group mean, and SE. (E) AZD4785 significantly inhibited tumor growth compared to PBS (TGI, 80%; P = 0.001) and CTRL ASO (TGI, 67%; P < 0.001). Data are shown as the geometric mean of the tumor volume and SE. (F) Surrogate endpoint survival graphs for the NCI-H1944 study. Data are shown as the percentage of remaining animals with tumors <4× the initial starting volume in each treatment group.

Fig. 5. PD and efficacy of AZD4785 in a KRAS mutant lung cancer PDX model.

(A) KRAS and DUSP6 mRNA expression was assessed by qRT-PCR in LXFA 983 cells in vitro after 72 hours treatment with AZD4785 or CTRL ASO. Representative data from a minimum of two independent experiments are shown. (B) Western blot analysis of MAPK signaling in LXFA 983 cells after 72 hours treatment with AZD4785 and CTRL ASO or 24 hours treatment with selumetinib. (C to G) Mice bearing LXFA 983 tumors were treated with PBS, AZD4785, or CTRL ASO at 50 mpk 5× weekly. (C) KRAS, DUSP4, and DUSP6 mRNA expression were measured in LXFA 983 tumors by qRT-PCR after 4 weeks of dosing. The expression was normalized to 18S rRNA and shown relative to PBS. Data are shown as individual tumor data, treatment group geometric mean, and SE. (D and E) KRAS protein expression measured by IHC in LXFA 983 tumors after 4 weeks of dosing. (D) Representative images are shown (scale bars, 100 μm) and (E) quantitation of KRAS (membrane H score) determined by Image analysis platform. Data are the mean membrane H score and SE. (F) AZD4785 significantly inhibited LXFA 983 tumor growth at the end of study compared to PBS (TGI, 98% at day 41, P < 0.001) and CTRL ASO (TGI, 95%; P < 0.001). Data are shown as the geometric mean of the tumor volume and SE. The arrow indicates the time point at which half of the animals were removed for PD analysis. (G) Surrogate endpoint survival graphs for the LXFA 983 study. Data are shown as the percentage of remaining animals in each treatment group with tumors <4× the initial starting volume. Two cohorts of animal are shown: data from all animals (n = 15) up to day 27 (filled symbol, unbroken line) and data from animals (n = 8) treated for 41 days (open symbol, broken line).

Patient-derived xenograft (PDX) models are thought to be more biologically relevant to human disease than standard cell line–derived xenograft models and reflect the histological characteristics of human tumors better (33). LXFA 983 is a lung cancer PDX model that carries a KRAS G12C mutation. In vitro studies with AZD4785 confirmed both free uptake of ASO, resulting in KRAS mRNA knockdown and KRAS dependency of MAPK pathway signaling in this model (Fig. 5, A and B). Systemic delivery of AZD4785 to LXFA 983 tumor-bearing mice reduced tumor expression of KRAS (84%; P < 0.001), DUSP4 (44%; P = 0.057), and DUSP6 (61%; P = 0.004) mRNA (Fig. 5C) and was furthermore associated with a reduction in membrane KRAS protein detected in tumors by IHC compared to controls (Fig. 5, D and E). PD effects of AZD4785 in the LXFA 983 PDX model correlated with 98% TGI (P < 0.001) at day 41 of dosing compared to PBS (Fig. 5F and fig. S10C). In the LXFA 983 model, some TGI was observed with the CTRL ASO (53%; P = 0.019) compared to the PBS (Fig. 5F). The reason for this is unclear but is not associated with either KRAS down-regulation or inhibition of downstream effector pathway activity (Fig. 5C). The AZD4785-treated animals showed 95% TGI (P < 0.001) when compared to the CTRL ASO group, and treatment with AZD4785 but not CTRL ASO delayed the time to reach the surrogate survival end point (Fig. 5G). We also assessed PD and efficacy of AZD4785 in a KRAS wild-type NSCLC PDX model, LXFA 526 (Fig. 6). Despite potent knockdown of KRAS in tumors [89% mRNA (P < 0.001) and 50% protein] (Fig. 6, A to C), this model showed less sensitivity to AZD4785 than the KRAS mutant models, with no significant difference in tumor growth compared to the CTRL ASO (Fig. 6D and fig. S10D) or impact on time to reach the surrogate survival end point (Fig. 6E).

Fig. 6. PD and efficacy of AZD4785 in a KRAS wild-type lung cancer PDX model.

Mice bearing LXFA 526 tumors were treated with PBS, AZD4785, or CTRL ASO at 50 mpk 5× weekly. (A) KRAS and DUSP6 expression measured by qRT-PCR in the LXFA 526 tumors after 4 weeks of dosing. The expression is normalized to 18S rRNA and expressed relative to PBS. (B and C) KRAS protein expression measured by IHC in LXFA 526 tumors after 4 weeks of dosing. (B) Representative images are shown (scale bars, 100 μm) and (C) quantitation of KRAS (H score) determined by Image analysis platform. Data are the mean H score and SE. (D) AZD4785 showed modest inhibition of tumor growth versus PBS (51%; P = 0.001); however, there was no significant TGI compared to the CTRL ASO. Data shown are treatment group geometric mean and SE. (E) Surrogate endpoint survival graphs for the LXFA 526 study. Data are shown as the percentage of remaining animals with tumors <4× the initial starting volume in each treatment group.

Tissue concentrations of ASO in preclinical models have been used to predict efficacious dose levels in humans (34). Measurement of AZD4785 tissue concentrations from the LXFA 983 studies, in which good target knockdown and TGI were observed, showed liver concentrations of AZD4785 ranging from 215 to 407 μg/g at the end of the studies (Table 1).

Table 1. AZD4785 PK measurements from liver at the end of the LXFA 983 studies.
View this table:

AZD4785 efficacy and target engagement were demonstrated in several additional KRAS mutant cell line–based and PDX in vivo models (table S3). Furthermore, additional KRAS cEt ASOs, with binding sites different from AZD4785 (fig. S11A), also showed PD and efficacy in vivo in the NCI-H358 model (fig. S11, B and C). These data as well as the phenotypic specificity of AZD4785 for KRAS mutant tumors are consistent with the notion that the antitumor activity of the KRAS ASOs is being driven by selective KRAS down-regulation. Together, these data demonstrate that systemic delivery of unformulated AZD4785 at clinically relevant doses in preclinical models of lung cancer can achieve KRAS knockdown and antitumor activity selectively in KRAS mutant tumors.

ASO-mediated systemic KRAS knockdown is well tolerated in mouse and monkey

AZD4785 targets both wild-type and mutant KRAS therefore, to better understand the therapeutic window, we assessed the physiological consequences of the ASO-mediated inhibition of KRAS in normal tissues. Because AZD4785 does not down-regulate murine KRAS, to do this, we identified potent and specific cEt ASOs targeting mouse KRAS mRNA and characterized their activities upon systemic delivery to adult mice. Mouse-specific KRAS ASOs administered by subcutaneous injections produced robust knockdown of KRAS mRNA and protein in multiple tissues, with selectivity over NRAS and HRAS (Fig. 7, A and B, fig. S12, and table S4). Knockdown of KRAS was not associated with any adverse effects (fig. S13 and tables S5 and S6) even at the highest doses, where strong KRAS mRNA knockdown was achieved in many tissues (94% in liver, 92% in lung, 81% in heart, and 73% in kidney). Although KRAS knockdown in the liver reached 94%, no notable liver-related tolerability signals were detected during both biomarker and histopathology assessment (Fig. 7, A and B, fig. S13, and tables S5 and S6), suggesting that although genetic disruption of KRAS is detrimental to the developing fetal liver (35), KRAS knockdown in this organ in adults is well tolerated. In addition, no impact on MAPK pathway signaling transcripts (DUSP4, DUSP6, or ETV4) was observed in the liver of the mice, supporting functional redundancy of the RAS isoforms in the adult tissue (fig. S12B). Changes that were observed in the clinical chemistry and clinical pathology profiles (modest changes in plasma concentrations of alanine transaminase and aspartate transaminase, liver weight, extramedullary hematopoiesis in the spleen, and minimal to mild basophilia of enlarged Kupffer cells in the liver) in KRAS ASO–treated mice were similar to those observed in the group treated with CTRL ASO and therefore were related to high ASO accumulation in the tissues and not to KRAS knockdown.

Fig. 7. ASO-mediated KRAS knockdown in mouse and monkey to assess tolerability.

(A and B) BALB/c mice were treated twice weekly with PBS, CTRL ASO, or mouse-selective KRAS (mKRAS) ASOs at 50 mpk for 6 weeks. (A) IHC analysis of KRAS protein in mouse tissues at the end of the study (scale bars, 200 μm). (B) qRT-PCR measuring KRAS mRNA in mouse tissues at the end of the study. mRNA expression was normalized to total RNA and expressed relative to PBS. Individual animal data, mean, and SE are shown. (C) Sequence alignment of monkey and human KRAS mRNA isoforms 3′ of the open reading frame, with the binding site of AZD4785 highlighted in red. (D) qRT-PCR demonstrating KRAS mRNA down-regulation in primary cynomolgus monkey hepatocytes after transfection with AZD4785. mRNA expression was normalized to total RNA and presented relative to PBS. Cynomolgus monkeys were treated for 6 weeks with AZD4785 or vehicle. For the first week, animals were subcutaneously dosed with 40 mpk four times and subsequently once weekly with 40 mpk. (E) IHC analysis of KRAS protein in monkey tissues at the end of the study (scale bars, 200 μm). (F) qRT-PCR measuring KRAS mRNA in monkey tissues at the end of the study. mRNA expression was normalized to total RNA and presented relative to PBS. Individual animal data, mean, and SE are shown.

The human drug candidate AZD4785 is perfectly matched to the cynomolgus monkey KRAS mRNA (Fig. 7C) and produced potent dose-dependent knockdown of KRAS mRNA in treated cynomolgus monkey hepatocytes in vitro (Fig. 7D). Therefore, we determined the tolerability of AZD4785 in male cynomolgus monkeys in a 6-week study at high (40 mpk/week) dose. Similar to the mouse-specific ASO, AZD4785 produced substantial and selective knockdown of KRAS mRNA and protein in multiple tissues (Fig. 7, E and F, and fig. S14), without any detectable adverse effects or histopathological changes, which could be attributed to KRAS inhibition (tables S7 to S11).

In toxicology studies, selective MEK1/2 inhibitors were reported to cause soft tissue and vascular mineralization in rats. This phenotype was preceded by 1,25-dihydroxyvitamin D3 and serum inorganic phosphorus (IP) increases (36). In toxicology studies with both mouse and monkey KRAS ASOs, we have not detected any changes in the circulatory concentrations of IP and calcium (fig. S15).

Together, these data demonstrate that systemic inhibition of KRAS with potent KRAS-specific cEt ASOs in the adult mouse or monkey was well tolerated. Minimal changes observed in some clinical chemistry parameters and clinical pathology profiles were consistent with reported effects of high ASO accumulation in the tissues (37) and were unrelated to the KRAS inhibition.

DISCUSSION

Targeted inhibition of KRAS has been a challenge in cancer research for more than 30 years. This is due to the high affinity of KRAS for GDP and GTP as well as the lack of well-defined druggable sites on the protein surface. However, some progress have been made in the development of covalent inhibitors that bind and inhibit the G12C mutant isoform of KRAS (15, 38). The second-generation inhibitor ARS-853 covalently interacts with KRAS G12C bound to GDP, preventing nucleotide exchange and locking it in an inactive state (16, 17). This compound has shown promising activity in KRAS G12C mutant cells in vitro, inhibiting proliferation in 3D assays and inhibiting the activity of downstream KRAS effector pathways (16, 17). However, the in vivo pharmacologic activity of ARS-853 in tumor models has yet to be assessed. Moreover, these inhibitors are at early stages of discovery and, as mentioned previously, will be limited to use in KRAS G12C mutant subpopulation of KRAS mutant cancers. Therapeutic siRNAs have also been explored to target a subpopulation of KRAS-dependent tumors. siG12D LODER is a KRAS G12D mutant–selective siRNA encapsulated in a biodegradable polymer delivery vehicle and has shown antitumor activity in mouse models of pancreatic cancer (18). This technology is currently in clinical trials for the treatment of pancreatic cancer (NCT01676259). However, systemic delivery of siRNA treatments required to treat metastatic tumors is still a challenge and would require improvements in lipid or nanoparticle technology (39).

AZD4785 is a first-in-class next-generation cEt ASO drug that selectively down-regulates KRAS mRNA and protein and is efficacious in preclinical KRAS mutant lung cancer models in vivo through unformulated systemic delivery. AZD4785 is selective for KRAS and does not down-regulate the closely related isoforms, NRAS or HRAS however, because it binds to the 3′UTR of KRAS, it is able to target all mutant isoforms of KRAS and therefore has a broad therapeutic potential across tumor types. Thus far, we have concentrated our studies with AZD4785 as a therapeutic drug in KRAS mutant NSCLC in vivo; however, AZD4785 has also demonstrated activity in KRAS mutant colon and pancreatic cancer cell lines in vitro. Antitumor activity in vitro and in vivo was not observed with AZD4785 in KRAS wild-type models, despite potent KRAS knockdown, consistent with the expected phenotypic activity of a selective KRAS targeting agent. Moreover, the doses resulting in robust efficacy are projected to be achievable in humans based on previous experience with cEt ASOs (24).

The sensitivity of individual KRAS mutant preclinical models to AZD4785 is dependent upon both the ability of the cells to take up ASO (25) and the inherent sensitivity of the tumor cells to KRAS depletion. This depends upon the genetic and gene expression context of the tumor with a KRAS mutation, such that tumors with high baseline expression of KRAS and epithelial markers may show the highest level of dependency (26).

The MAPK cascade (RAF-MEK-ERK) is a key effector pathway regulated by KRAS, and genetic loss of CRAF, MEK1/2, and ERK1/2 prevents mutant KRAS-driven lung tumors (40, 41). Our data demonstrate that inhibition of KRAS is different from MAPK pathway inhibition by MEK1/2 or ERK1/2 inhibitors in several ways. First, although AZD4785 down-regulates both wild-type and mutant KRAS, the robust inhibition of cell proliferation and downstream signaling is selective to KRAS mutant tumor cells. In contrast, the activities of selumetinib and SCH772984 were not dependent upon the presence of a KRAS mutation, potentially limiting the therapeutic window of these drugs. Second, because of the relief of negative-feedback controls, selumetinib and SCH772984 increase MEK1/2 phosphorylation at S218/222, which can cause pathway reactivation and may be a mechanism of clinical resistance to these drugs (42, 43). In contrast, down-regulation of KRAS by AZD4785 did not increase phosphorylation of MEK1/2 and could limit selumetinib-induced pathway reactivation in some KRAS mutant cells. This suggests that KRAS is a key node in pathway reactivation and highlights an attractive drug combination opportunity for AZD4785 and MAPK pathway inhibitors. The combination of AZD4785 and selumetinib showed enhanced inhibition of KRAS mutant cell proliferation over either drug given as monotherapy. Finally, whereas selumetinib and SCH772984 in some cell backgrounds increased pAKT, likely through relief of feedback of upstream receptors (44), this was not observed with AZD4785, and in fact, certain cell lines showed down-regulation of pAKT, consistent with the PI3K/AKT cascade being another effector pathway of RAS (45, 46).

An important question for any KRAS-targeted therapeutic agent is whether these need to be selective for the mutant form of KRAS or whether nonselective inhibition of both the mutant and wild-type forms of KRAS will have the desired antitumor activity and the required safety profile to be valuable cancer drugs. With the KRAS ASO approach described here, we demonstrate that simultaneous depletion of both mutant and wild-type KRAS results in selective antitumor activity in cells expressing mutant KRAS. With respect to the potential safety of KRAS depletion in normal cells and tissues, gene disruption experiments have demonstrated that wild-type KRAS function is essential for normal mouse early development because KRAS homozygous null embryos die between days 12 and 14 of gestation and exhibit fetal liver defects and severe anemia (35). However, there are no reported studies addressing the continued requirement for KRAS in adult organisms. Moreover, the tolerability of therapeutic inhibition of KRAS in adult organisms will depend on the PK and distribution properties of the therapeutic modality used. To assess the tolerability of ASO-mediated KRAS depletion in adult animals, we identified and evaluated potent mouse-specific cEt KRAS ASOs. Systemic delivery of mouse KRAS ASOs produced robust target knockdown of KRAS mRNA and protein in normal mouse tissues, including the liver, lung, heart, and kidney. Encouragingly, for the therapeutic approach described here, despite the robust knockdown in tissues, no significant target-related tolerability issues were observed. Furthermore, because of complete sequence homology, AZD4785 inhibits monkey as well as human KRAS and was thus evaluated in a 6-week monkey safety study. Similar to the findings in mice, ASO-mediated knockdown of KRAS in cynomolgus monkey was confirmed in normal tissue and was well tolerated. These data suggest that KRAS function may be compensated for by NRAS and/or HRAS isoforms in adult tissues where ASOs are active and supports a therapeutic window for AZD4785-mediated KRAS depletion, because it had no activity on either NRAS or HRAS mRNA or protein. We also observed a differentiation in tolerability profiles for specific KRAS ASOs over MEK inhibitors, with no observed increases in plasma phosphates or calcium, which have been reported to precede tissue mineralization induced by MEK inhibitors (36). This suggests potential redundancy in control of MEK activity by different RAS isoforms in normal tissue, which is further supported by our observation that KRAS knockdown in murine and monkey liver has little impact on known MAPK pathway regulated transcripts, and highlights a differential tolerability profile for AZD4785.

Therapeutic ASOs are becoming an increasingly attractive therapeutic modality to target traditionally difficult drug pathways. Increased understanding of biological mechanisms involved in tissue distribution, cellular uptake, and intracellular trafficking of oligonucleotides has enabled new approaches to enhance the delivery and activity of oligonucleotides. These include ligand-oligonucleotide conjugates [such as N-acetylgalactosamine conjugation for effective delivery to hepatocytes (47)], lipid- and polymer-based nanoparticles (48, 49), targeted delivery with antibody conjugates (50), and the combination with small molecules that enhance uptake and activity of oligonucleotides (51). Future strategies building off of these approaches will be important to maximize the potential of therapeutic ASOs to broadly treat human diseases.

This study has been limited to assessing the antitumor activity of AZD4785 in subcutaneous xenograft and PDX models of lung cancer in immunocompromised mouse models. Thus, we have yet to explore the impact of a therapeutic KRAS ASO on KRAS mutant tumors in situ or in the presence of an intact immune microenvironment. In conclusion, our data demonstrate that AZD4785 is a potent and well-tolerated KRAS cEt ASO with robust antitumor activity at doses relevant to the clinical setting and suggest that AZD4785 has potential as a therapeutic to help address the high unmet clinical need represented by mutant KRAS-driven human cancers.

MATERIALS AND METHODS

Study design

The overall objectives of this study were to assess the in vitro and in vivo activity of AZD4785, a generation 2.5 cEt-modified ASO targeting KRAS mRNA, in KRAS mutant and wild-type preclinical models, and to assess the tolerability of KRAS down-regulation by ASO treatment in normal tissues.

The activity and selectivity of AZD4785 were first evaluated in multiple KRAS wild-type and mutant cell lines, assessing effects on colony formation in soft agar and signaling end points known to be modulated by KRAS activity. In vitro experiments were not performed blinded and were measured in technical replicates, with a minimum of two biological replicates experiments performed.

Next, the in vivo activity of systemically delivered unformulated AZD4785 was studied in multiple KRAS mutant cell line–based and patient-derived subcutaneous xenograft tumors, comparing the antitumor activity to vehicle and control ASO. Antitumor activity was assessed by measuring tumor volume, KRAS expression knockdown, and the activity of KRAS-regulated downstream effector pathways. Age- and gender-matched animals were randomly assigned into treatment groups (n = 6 to 15) to ensure an equal tumor size (100 to 200 mm3) across groups at the initiation of the study. In vivo studies were not performed blinded. Experimental replicates for in vivo studies were as follows: NCI-H358, four replicates; LXFA 983, two replicates; and LXFA 526 and NCI-H1944, one replicate.

The tolerability of knocking down KRAS expression by an ASO was tested in age-matched male and female mice, using mouse-specific KRAS ASO, or in male cynomolgus monkeys, using AZD4785. Tolerability was assessed by measuring KRAS expression knockdown in multiple normal tissues and measuring changes in clinical chemistry and clinical pathology profiles compared to vehicle and control ASO-treated animals. Microscopic evaluation of tissue sections was performed by a pathologist, and findings were peer-reviewed by separate pathologists. No data points were identified and removed as outliers.

Statistics

Tumor growth inhibition (%TGI) from the start of treatment was assessed by comparison of the geometric mean change in tumor volume for the control and treated groups. Statistical significance was evaluated using a one-tailed t test. For PD analysis, statistically significant changes were determined using analysis of variance (ANOVA).

Additional materials and methods are available in the Supplementary Materials.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/9/394/eaal5253/DC1

Materials and Methods

Fig. S1. Effect of AZD4785 on KRAS, DUSP6, and ETV4 mRNA expression in KRAS mutant and KRAS wild-type cell lines.

Fig. S2. Sensitivity of NSCLC lines to KRAS knockdown by ASO in 2D versus 3D growth assays.

Fig. S3. Effect of AZD4785 on colony formation in KRAS mutant and KRAS wild-type cell lines.

Fig. S4. Effect of AZD4785, selumetinib, and SCH772984 on the proliferation of KRAS mutant and KRAS wild-type cell lines.

Fig. S5. Effect of AZD4785, selumetinib, and SCH772984 on signaling in KRAS mutant and KRAS wild-type cell lines.

Fig. S6. Effects of AZD4785 and selumetinib combination on signaling and proliferation of NSCLC cells.

Fig. S7. In vivo study assessing the kinetics of tumor PD and liver PK of AZD4785.

Fig. S8. Effect of AZD4785 treatment on HRAS and NRAS expression in the NCI-H358 xenograft model.

Fig. S9. Tolerability of AZD4785 treatment in xenograft studies.

Fig. S10. Waterfall plots of xenograft and PDX studies.

Fig. S11. PD and efficacy of AZD4785 and additional cEt KRAS ASOs in the NCI-H358 model.

Fig. S12. Selectivity of murine KRAS ASOs and impact on MAPK pathway signaling in liver tissue from mice.

Fig. S13. Summary of the effects of the murine KRAS ASOs on body weight, plasma chemistry, and circulating blood cells in mice.

Fig. S14. Selectivity of AZD4785 and impact on MAPK pathway signaling in liver tissue from monkeys.

Fig. S15. Impact of murine KRAS ASOs on plasma concentrations of inorganic phosphates and calcium.

Table S1. Summary of predicted off-targets for AZD4785.

Table S2. Details of the cell lines used in the study.

Table S3. Activity of AZD4785 in KRAS mutant NSCLC xenograft models.

Table S4. Inhibition of KRAS mRNA across a panel of normal tissues after treatment of mice with murine KRAS ASOs.

Table S5. Summary of tissue histopathology in the mouse tolerability study after treatment with the murine-selective KRAS ASOs.

Table S6. Summary of body and organ weights, plasma chemistries, and circulating blood cells in the mice after treatment with the murine-selective KRAS ASOs.

Table S7. Summary of tissue histopathology in the monkey tolerability study after treatment with AZD4785.

Table S8. Summary of body and organ weights in the monkey tolerability study after AZD4785 treatment.

Table S9. Summary of circulating blood cells in the monkey tolerability study after AZD4785 treatment.

Table S10. Summary of plasma chemistries in the monkey tolerability study after AZD4785 treatment.

Table S11. Summary of urinalysis in the monkey tolerability study after AZD4785 treatment.

Table S12. Details of the parameters used for analyzing colony formation across the cell line panel.

Table S13. Summary of the individual animal tumor volumes in the xenograft and PDX studies.

References (52, 53)

REFERENCES AND NOTES

Acknowledgments: We thank S. Freier, A. Watt, D. Gattis, and M. Bell (Ionis Pharmaceuticals) for the design and in vitro screening of the human and mouse KRAS ASOs; L. Hettrick, C. May (Ionis Pharmaceuticals), and H. Musgrove (AstraZeneca) for the technical assistance with rodent studies; Y. Jiang (Ionis Pharmaceuticals) for the immunohistological procedures; the Korean Institute of Toxicology, T.-W. Kim, and S. Burel (Ionis Pharmaceuticals) for conducting monkey tolerability study; Oncotest for conducting PDX studies and providing PDX cell lines; A. Hiraide (Preclinical Sciences R&D, AstraZeneca) for PC9 cells; and G. Duchesne (Institute of Cancer Research) for HX 147 cells. Funding: The study was funded by AstraZeneca and Ionis Pharmaceuticals. Author contributions: S.J.R. and A.S.R. were involved in conception and design, acquisition, analysis and interpretation of data, study supervision, and writing of the manuscript. L.L.H. was involved in design, acquisition, analysis, and interpretation of some in vivo studies. S.K.P. was involved in acquisition and analysis of some in vivo studies. R.E. and A.S. were involved in acquisition and analysis of RT-PCR data from in vivo study samples. M.R. was involved in acquisition and analysis of IHC data from in vivo study samples. N.W. was involved in acquisition and analysis of in vitro combination data. L.K.B. and S.K.K. were involved in design, analysis, and interpretation of safety study data. C.R., K.H., M.Z., and D.C.B. were involved in concept and design of studies and reviewing of the manuscript. B.P.M. was involved in concept and design of studies and study supervision. P.D.L. and A.R.M. were involved in concept and design of studies, study supervision, and writing of the manuscript. Competing interests: S.J.R., L.L.H., R.E., A.S., N.W., M.R., C.R., L.K.B., S.K.K., K.H., M.Z., D.C.B., and P.D.L. were/are employees and shareholders of AstraZeneca. A.S.R., S.K.P., B.P.M., and A.R.M. are employees and shareholders of Ionis Pharmaceuticals. A.R.M. and A.S.R. are inventors on a patent application WO/2017/053722 that covers AZD4785 and is held/submitted by Ionis Pharmaceuticals. Data and materials availability: Researchers may obtain AZD4785 with a material transfer agreement from AstraZeneca and Ionis Pharmaceuticals. All reasonable requests for collaboration involving materials used in the research will be fulfilled provided that a written agreement is executed in advance between AstraZeneca and the requester (and his or her affiliated institution). Such inquiries or requests should be directed to the corresponding authors.
View Abstract

Navigate This Article